基于视频机器分析的目的地形象差异对比——以北京You Tube视频为例

Comparison of Destination Images Based on Video Analysis through Machine Learning——A Case Study on YouTube Videos of Beijing

大数据背景下,视频越来越成为用户获取目的地信息并建构旅游形象的主要渠道,鉴于用户生成视频(UGC)相较于目的地营销组织发布视频(OGC)更受欢迎,比较二者建构目的地形象的差异具有一定的理论意义及实践价值。然而到目前为止,旅游领域对视频内容的研究还较初级,且多采用人工分析方法,局限明显。近年来,计算机视觉领域在视频内容分析方面取得了巨大进展,基于此,文章采用计算机学科相关视频分析方法对YouTube网站上国外游客和北京市文化和旅游局发布的北京旅游视频展开研究。研究发现:第一,游客和目的地营销机构(DMO)借以建构目的地形象的属性主要是“人物”“文化艺术”和“基础设施”,在“交通”和“动植物”属性上表现较弱;第二,视频和文本素材在表征目的地“旅游景点”和“人物”属性上具有一致性,而“基础设施”属性在视频素材中比在文本中展现更充分;第三,UGC视频倾向表征北京“文化艺术”属性,尤其青睐表征北京建筑及文化场景,而DMO倾向表征更全面、宏观的北京形象,尤其突出展现北京的人物属性。

Destination image(DI)has been a popular topic of research in the field of tourism since it contributes majorly to the destination choice of tourists.Content created by tourists and published on the Internet based on their daily experiences is called“user-generated content”(UGC)while official content released by destination marketing organizations(DMOs)is termed“occupationally-generated content”(OGC).At present,although the comparative research between UGC and OGC has received wide attention from scholars,the research mainly focuses on DIs projected through texts and photos.Nowadays,as the 4G/5G communication networks develop,online videos have increasingly become the powerful channel for potential tourists to collect information about destinations and construct the DI.Dynamic and visual,videos contain richer information than photos and texts and can tell stories through rich narratives to make audiences further immersed in the destination.More and more potential tourists learn about destinations through travel videos and then make decisions.This means that DI is increasingly affected by online videos,a powerful medium for tourism marketing.Therefore,taking travel videos on YouTube as the research object,this paper explored how videos represent DIs in an age of rich media.This paper takes travel videos posted by international tourists on YouTube as UGC materials and those posted on“Visiting Beijing”,the official YouTube account of the Beijing Municipal Culture and Tourism Bureau,as OGC materials.The study of video content in the tourism field is still at an early stage and mostly adopts a manual analysis method with significant limitations.Video content analysis in the computer vision field has seen great progress in recent years.Thus,this paper introduces the deep learning framework and used video analysis methods such as scene detection and video captioning to deeply analyze the content and captions of UGC and OGC videos.In addition,this paper also compares the destination images displayed by different media(text and video)in view of the machine-generated Beijing DI.Finally,a comparative analysis is conducted to identify the differences in DI projection and scene design between the UGC and OGC videos.The study finds that:firstly,tourists and DMOs pay more attention to people,culture and art,and infrastructure in their DI construction and less attention to traffic and plants and animals.Secondly,there is not much difference in the representation of tourist attractions and people between texts and videos while the infrastructure of Beijing is fully demonstrated through videos rather than texts.Thirdly,UGC videos tend to show Beijing from the perspective of culture and art,especially the architecture and cultural scenes of Beijing,while the OGC videos of DMOs prefer to showcase Beijing across the board and at a macro level,highlighting,in particular,the local people.